I think that, in all the cases presented above something else is going on. It seems that decision makers generally do not process all the available information before reaching a conclusion. Rather, they draw a ‘tentative’ conclusion based on the in- formation that is close at hand. When explicitly asked to elaborate, when relating decisions to themselves, or when shaken up by the encounter of unusual situations, decision makers no longer rashly accept the first conclusion that springs to mind. Rather, they are inclined to take into account less readily available information. This may induce them to overrule their first conclusion and make a different de- cision.
One experiment that can be taken to support this claim shows that when agents are given full information at once, no framing effect is observed. For example, when the descriptions of the Asian disease programmes stress both the amounts of lives saved and lost, and hence both positive and negative information and associ- ations are at hand, the results are more or less the same as group 2 in Takemura’s elaboration experiments. Furthermore, there is no difference between frames in which the number of lives saved is put first and frames in which the amount of deaths is put first (Kühberger 1995, p.234).
A similar result is obtained by Sniderman and Theriault (2004) in the context of topic framing. They show that when test subjects are presented with two competing
frames, such as a frame stressing the impact of a new law on individual freedom and a frame stressing the impact of this law on public safety, the framing effects mutually cancel out (Sniderman and Theriault 2004, pp.153-156).
Putting together the insights gained so far, I claim that the underlying decision process of the framing effect can be described as follows. As argued in section 2.1.2, the use of specific words or formulations in a specific context gives rise to various associations. Some of these associations are triggered by the ‘objective’ descriptive content of the proposition, others are grounded in semantic properties of the words used, are the result of patterns or regularities one is implicitly aware of and which are tied to specific formulations, or arise due to the context in which specific words are used.
I have called these associations the ‘information’ (in the broadest sense of the word) contained by a frame. What happens in framing is that some of these as- sociations become more focal than others. That is, the use of a specific formula- tion triggers some associations right away. Other, less apparent information that can also be extracted from the decision problem requires more cognitive effort to be considered. Since the ‘immediate’ associations are often sufficient to reach a tentative conclusion, our decision process commonly ends before all the available information has been considered.
Some subjects, however, are not satisfied with this conclusion and can decide to take more information into account. This information is less focal and therefore requires more cognitive effort to process. Several factors can induce an agent to indeed make this effort. In this section, a few of them have been discussed: when one is asked to take more time (e.g., by providing a justification), when the problem gets personal, or when the mind is ‘shaken up’ by an unusual or extreme situation. However, it stands to reason that other, less direct factors play a role as well, such as the subject’s intelligence, mood, vigilance etc. This can also explain the lack of unanimity. While most subjects stop the decision process as soon as they have reached a first conclusion on the target proposition, some others proceed and take more information into account.
It is important to note that this process of considering information is the same for both topic framing and valence framing. Successful topic framing causes in- formation about a certainaspect of a problem to be processed first, and thereby often leads to a conclusion largely based on information pertaining to this aspect (e.g., economic benefits). Valence framing causespositive or negativeinformation related to a problem to be considered first, thereby often leading to a conclusion
based on this positive or negative information. In both cases, the same underlying process of giving priority to the primed information is at work.
Hence, the notion of ‘partial information’ can be said to be at the heart of the framing effect. A frame greatly influences the order in which the information con- tained by a decision problem is processed (i.e., which associations are triggered immediately and which require more cognitive effort), and thereby influences the information used to reach a first (and often final) decision.
In a way, this idea of partial information is reminiscent of a central claim put forward by Daniel Kahneman in his best-selling book Thinking, Fast and Slow
(2011). Kahneman argues here that human beings have two systems or modes for decision making. There is a ‘hasty’ mode, in which we use various heuristics and biases to quickly decide on an issue. As a consequence, some information may be lost or tarnished in this process. Secondly, there is a ‘slow’ mode, in which we take more time to carefully consider all available information. Prospect Theory is designed to model this first mode, while the second one much more resembles classical logic.
In the next chapter, I will shown that it need not be the case that human beings indeed use two separate systems for decision making. I will present one and the same model that can accommodate both partial and full information processing.
In sum, one can say that the underlying decision process of the framing ef- fect has three major tenets; (1) the central role played by a variety of associations or information extracted from a decision problem (not merely confined to ‘risk’, ‘reference points’ or ‘good news’); (2) the varying degrees of accessibility of this information, chiefly dependent on the frame that is used and the cognitive effort it takes to access (implicit) information; and (3) the resulting consequence that we often only take partial information into account when making decisions, as this is often enough to reach a (tentative) conclusion.
It is not hard to see that Gold and List’s notion of path dependence fits well with these tenets (especially 2 and 3). Therefore, it will play a key role in the model presented in chapter 3. However, despite the fruitfulness of this notion, I have also argued that the model provided by Gold and List themselves is too static and narrow to provide a full account of all facets of the framing effect. As it turned out, the model faces problems in representing goal framing, and it cannot provide a proper explanation of the results discussed in the previous section, such as the elaboration effect. The model lacks a flexible mechanism for looking beyond our first, preliminary, conclusion and for considering less focal information.
Apart from this, there is another problematic aspect of LPD that has not been discussed so far: the alleged violation of the standards for rationality that Gold and List take to be necessary for the framing effect to occur. I will turn to that now.